UR Scholarship Repository
Law Faculty Publications
School of Law
2018
Punishing Risk
Erin Collins
University of Richmond - School of Law, [email protected]
Follow this and additional works at:
https://scholarship.richmond.edu/law-faculty-publications
Part of the
Criminal Law Commons
, and the
Criminal Procedure Commons
This Article is brought to you for free and open access by the School of Law at UR Scholarship Repository. It has been accepted for inclusion in Law Faculty Publications by an authorized administrator of UR Scholarship Repository. For more information, please contact
Recommended Citation
ERIN COLLINS*
Actuarial recidivism risk assessments-statistical predictions of the likelihood of future criminal behavior-drive a number of core criminal justice decisions, including where to police, whom to release on bail, and
how to manage correctional institutions. Recently, this predictive approach to criminal justice entered a new arena: sentencing. Actuarial sentencing has quickly gained a number of prominent supporters and is being implemented across the country. This enthusiasm is understand-able. Its proponents promise that actuarial data will refine sentencing decisions, increase rehabilitation, and reduce reliance on incarceration. Yet, in the rush to embrace actuarial sentencing, scholars and policy makers have overlooked a crucial point: actuarial risk assessment tools are not intended for use at sentencing. In fact, their creators explicitly warn that these tools were not designed to aid decisions about the length of a sentence or whether to incarcerate someone. Nevertheless, that is precisely how those who endorse actuarial sentencing-including the American Law Institute in the recently revised Model Penal Code for Sentencing-suggest they should be used.
Actuarial sentencing is, in short, an unintended, "off-label" applica-tion of actuarial risk informaapplica-tion. This Article reexamines the promises of actuarial sentencing in light of this observation and argues that it may cause a number of equally unintended and detrimental consequences. Specifically, it contends that this practice distorts, rather than refines, sen-tencing decisions. Moreover, it may increase reliance on incarceration-and it may do so for reasons that undermine the fairness incarceration-and integrity of the criminal justice system.
TABLE OF CONTENTS
INTRODUCTION ... 58
I. THE RISE OF ACTUARIAL SENTENCING ... 63
A. THE PRACTICE OF ACTUARIAL SENTENCING ... 63
1. Actuarial Risk Assessment ... 63
* Assistant Professor, University of Richmond School of Law. C 2018, Erin Collins. Thanks to
Russell Gold, Jessica Eaglin, Jessica Erickson, Corinna Lain, Dawinder Sidhu, Jocelyn Simonson, Allison Tait, and the participants of the 2017 American Association of Law Schools Midyear Meeting Workshop, the 2017 Mid-Atlantic Junior Faculty Forum, the 2017 Southeastern Association of Law Schools Meeting, and the University of Richmond School of Law faculty workshop for helpful comments and conversations.
2. Sentencing Actuarially ... 66
a. Sentence Length Decisions ... 67
b. Sentence Location Decisions ... 69
B. THE PROMISES OF ACTUARIAL SENTENCING ... 72
1. Reducing Recidivism Through Rehabilitation ... 72
2. Refining Risk Prediction ... 74
3. Increasing Efficiency ... 76
II. THE CORRECTIONAL ORIGINS OF ACTUARIAL SENTENCING ... 77
A. CORRECTIONAL RISK ASSESSMENT: ASSESSING RISK TO ADMINISTER PUNISHMENT ... 78
1. Paroling Risk ... 78
2. Correcting Risk ... 79
3. Correctional Risk Assessment Principles: Use Only as D irected ... 83
B. ACTUARIAL SENTENCING: ASSESSING RISK TO DETERMINE PUNISHMENT ... 85
1. Going "Off-Label" in Practice: Sentencing Risk ... 85
2. Going "Off-Label" in Theory: Incapacitating Risk ... 87
III. THE PROBLEMS OF ACTUARIAL SENThNCING ... 91
A. QUESTIONING THE BENEFIT ... 91
B. IDENTIFYING "OFF-LABEL" COSTS ... 98
C. REASSESSING ACTUARIAL SENTENCING ... 106
C ONCLUSION ... 108
INTRODUCTION
Predictive data about future criminal activity increasingly determine an indi-vidual' s fate within the criminal justice system. Data may determine if a person is arrested and charged. If he is, that data will influence whether he is released on bail. And if he is then convicted and incarcerated, data may determine where he is housed, what programs he must complete, and whether he is eventually released on parole. And now, as a result of a practice that is gaining popularity, predictive data may even determine the length of his sentence and whether he serves that sentence in his own home or in prison.
Actuarial sentencing, also called "evidence-based sentencing,"' is the practice of using actuarial risk assessment tools to guide sentencing decisions. Actuarial risk assessment tools seek to predict the likelihood an individual will commit crimes in the future based on the presence or absence of factors that statistically correlate with recidivism.2 Proponents claim that judges are already making these predictions, and actuarial tools simply help them do it better.3 In other words, they imply that this use of risk assessment information refines and enhances an established, normatively sound practice but does not fundamentally change the role of risk prediction at sentencing.
Actuarial sentencing is part of a larger movement that aims to be smart, rather than tough, on crime.4 This movement embraces data-driven practices to ensure that prison is reserved for those who represent a true danger to the public.5 Proponents promise that actuarial sentencing and other evidence-based practices will help us reorient the criminal justice system away from backward-looking, retributivist considerations and toward the future, with greater emphasis on increasing public safety through rehabilitation.6 In theory, at least, this new
"brand" of criminal justice uses data to help identify who can serve their sentence
1. See Jessica M. Eaglin, Constructing Recidivism Risk, 67 EMORY L.J. 59, 61-62 (2017); Sonja B. Starr, Evidence-Based Sentencing and the Scientific Rationalization of Discrimination, 66 STAN. L. REV.
803, 809 (2014) (describing "evidence-based sentencing"); see also NAT'L CTR. FOR STATE COURTS, NCSC FACT SHEET: EVIDENCE-BASED SENTENCING 1 (Aug. 2014), http://www.ncsc.org/-/media/ Microsites/Files/CSI/EBS%20Fact%20Sheet%208-27-14.ashx [https://perma.cc/3L6N-DQX6] (same). Other scholars have called this practice "risk-based sentencing."
2. See MODEL PENAL CODE: SENTENCING § 6B.09 rptrs. note a (AM. LAW INST., Proposed Final Draft 2017).
3. See, e.g., Jordan M. Hyatt et al., Follow the Evidence: Integrate Risk Assessment into Sentencing, 23 FED. SENT'G REP. 266, 266 (2011) (arguing that introducing risk assessment information into sentencing "would hardly represent a sea change" because "[w]ith varying degrees of formality, judges already consider risk at sentencing"); Richard P. Kern & Mark H. Bergstrom, A View from the Field: Practitioners' Response to Actuarial Sentencing: An "Unsettled" Proposition, 25 FED. SENT'G REP.
185, 185 (2013) ("[R]isk has and will continue to be used by courts at sentencing, whether formally or informally.").
4. See generally J.C. Oleson, Risk in Sentencing: Constitutionally Suspect Variables and Evidence-Based Sentencing, 64 SMU L. REV. 1329, 1336-37 (2011) (identifying the interest in actuarial sentencing as part of the call to "sentence 'smarter ); TRACY W. PETERS & ROGER K. WARREN, NAT'L CTR. FOR STATE COURTS, GETTING SMARTER ABOUT SENTENCING: NCSC's SENTENCING REFORM SURVEY (2006), http://www.ncsc.org/-/media/ Microsites/Files/CSI/GettingSmarter SentencingReformSurveyFinalPub. ashx [https://perma.cc/4XDW-AWCV] (identifying evidence-based practices as a "smart" sentencing reform).
5. See, e.g., Hon. Michael Marcus, Smarter Sentencing: On the Need to Consider Crime Reduction as a Goal, 40 CT. REV. 16, 22 (2004).
6. See, e.g., Kern & Bergstrom, supra note 3, at 185 ("[T]he consideration of risk does represent a shift in the purposes of sentencing, moving from backward-looking retributive approach with a focus on uniformity, proportionality, and reduction of unwarranted disparity to a forward-looking utilitarian approach with a focus on public safety and crime reduction."); see also Peggy McGarry, Introduction to RAM SUBRAMANIAN ET AL., VERA INST. OF JUSTICE CTR. ON SENTENCING & CORR., RECALIBRATING
JUSTICE: A REVIEW OF 2013 STATE SENTENCING AND CORRECTIONS TRENDS 2 (2014), https://www.vera. org/publications/state- sentencing-and-corrections -trends -2013 [https://perma.cc/N2DC-DQ6L] ("From appalling incarceration numbers, budgetary crises, and greater public knowledge, this momentum for reform has redirected the discussion on crime away from the question of how best to punish to how best to achieve long-term public safety.").
in the community without threatening the public and, in so doing, conserves ex-pensive incarceratory resources for those who cannot.7
In light of these promises, it is unsurprising that actuarial sentencing has been met with great acclaim. It has been embraced by the National Center for State Courts8 and the Conference of Chief Justices and the Conference of State Court Administrators.9 Present and former state sentencing commissioners have endorsed it,"0 and policymakers have called it the "most promising way forward" and "the new frontier in sentencing policy and practice."1 Actuarial sentencing is currently practiced in more than twenty jurisdictions,12 and its popularity is likely to grow in light of a recent notable development: the American Law Institute adopted the practice in its 2017 revision of the Model Penal Code for Sentencing.1 3
Recently, however, some scholars have begun to caution against this practice. They have argued that the risk tools incorporate certain characteristics-such as gender and socioeconomic status-that raise constitutional concerns.1 4 They have questioned the tools' accuracy15 and their purported objectivity.1 6 And they have highlighted that this practice, which involves making predictions based on group data, undermines the principle that sentences should be
7. See WILLIAM R. KELLY WITH ROBERT PITMAN & WILLIAM STREUSAND, FROM RETRIBUTION TO PUBLIC SAFETY: DISRUPTIVE INNOVATION OF AMERICAN CRIMINAL JUSTICE 177 (2017) (calling for a new "brand" of the criminal justice system that "involves problem solving and recidivism reduction").
8. See PAMELA M. CASEY ET AL., NAT'L CTR. FOR STATE COURTS, USING OFFENDER RISK AND NEEDS ASSESSMENT AT SENTENCING: GUIDANCE FOR COURTS FROM A NATIONAL WORKING GROUP 7 (2011), http://www.ncsc.org/-/media/microsites/files/csi/rna%20guide%20final.ashx [https://perma.ccIW9QC-LHCD].
9. See CONFERENCE OF CHIEF JUSTICES & CONFERENCE OF STATE COURT ADM'RS, RESOLUTION 7: IN SUPPORT OF THE GUIDING PRINCIPLES ON USING RISK AND NEEDS ASSESSMENT INFORMATION IN THE SENTENCING PROCESS (2011), http://www.ncsc.org/-/media/Microsites/Files/CSI/RNA%202015/Support% 20Guiding%20Principles%20Using%20RNAs.ashx [https://perma.cc/NC86-3YD7].
10. See Kern & Bergstrom, supra note 3, at 185. Richard Kern was the Director of the Virginia Criminal Sentencing Commission and Mark Bergstrom is the Executive Director of the Pennsylvania Commission on Sentencing.
11. See Roger K. Warren, Evidence-Based Sentencing: Are We Up to the Task?, 23 FED. SENT'G REP. 153, 153, 157 nn.1 & 3 (2010).
12. See Starr, supra note 1, at 809.
13. MODEL PENAL CODE: SENTENCING § 6B.09 (AM. LAW INST., Proposed Final Draft 2017) (the section is entitled Evidence-Based Sentencing; Offender Treatment Needs and Risk of Reoffending). The Model Penal Code is highly influential upon state practices and its "code and its commentaries have been the intellectual focus of much American criminal law scholarship." Paul H. Robinson & Markus D. Dubber, The American Model Penal Code: A Brief Overview, 10 NEW CRIM. L. REV. 319, 320 (2007).
14. See Starr, supra 1, at 806.
15. For example, computer scientists recently examined the reliability of a popular risk assessment tool and found that it "is no more accurate or fair than the predictions of people with little to no criminal justice expertise." Julia Dressel & Hany Farid, The Accuracy, Fairness, and Limits of Predicting Recidivism, 4 ScI. ADVANCES 1, 3 (2018); see also Brian Netter, Using Group Statistics to Sentence Individual Criminals: An Ethical and Statistical Critique of the Virginia Risk Assessment Program, 97 J.
CRIM. L. & CRIMINOLOGY 699,712 (2007).
16. See Eaglin, supra note 1, at 64 (arguing that "actuarial risk tools, while 'scientific' in the sense that developers use technology to assess risk, reflect normative judgments familiar to sentencing law and policy debates").
individualized.17 However, even the scholars who have begun to critically scrutinize actuarial sentencing have focused primarily on issues with the actua-rial method of assessment. Existing scholarship has thus failed to ask a critical question: is the risk inquiry these tools advance normatively sound?18
As a result, these conversations about the promises and perils of actuarial sen-tencing have overlooked an important point: actuarial risk assessment tools are not intended for use in sentencing.19 Rather, they were developed to guide decisions of correctional authorities-such as how to efficiently and effectively rehabilitate inmates during their incarceration-after a judge has announced a sentence.20 In other words, such tools were created to guide decisions about how to administer punishment, not about how much punishment is due. In fact, the social scientists who developed the tools that are being incorporated into sentenc-ing decisions expressly disavow their use to "assist in establishsentenc-ing the just pen-alty," specifically in decisions about whether to incarcerate and the length of the sentence.21 And yet, that is precisely how they are being used. Actuarial sentenc-ing is, in short, an "off label" application of these predictive tools.22
17. See Dawinder S. Sidhu, Moneyball Sentencing, 56 B.C. L. REV. 671,702-04 (2015).
18. See, e.g., Starr, supra note 1, at 872 (concluding that "[r]isk prediction is here to stay as part of sentencing, and perhaps actuarial instruments can play a legitimate role," but suggesting these instruments "should not include demographic and socioeconomic variables"); see also Sidhu, supra note 17, at 687 (noting that "forecasting the risk of recidivism is an important function in the criminal justice system").
19. See BERNARD E. HARCOURT, AGAINST PREDICTION: PROFILING, POLICING, AND PUNISHING IN AN ACTUARIAL AGE 188 (2007) (noting that actuarial prediction instruments "were generated, created, driven by sociology and criminology. .. .They had no root, nor any relation to the jurisprudential theories of just punishment"). One exception to this general oversight is a recent white paper that notes the possibility that judges may use a risk tool that was "designed for a correctional-not a sentencing-population." Jordan M. Hyatt & Steven L. Chanenson, The Use of Risk Assessment at Sentencing: Implications for Research and Policy 8 (Villanova Law Sch. Pub. Law & Legal Theory Working Paper Series, Working Paper No. 2017-1040, 2016), https://digitalcommons.law.villanova.edu/cgi/viewcontent. cgi?referer-https://www.google.com/&httpsredir- 1 &article= 1201 &context=wps
[https://perma.cc/SXW5-QZQ5].
20. HARCOURT, supra note 19, at 187-88.
21. See Malenchik v. State, 928 N.E.2d 564, 567, 572 (Ind. 2010) (quoting D.A. ANDREWS & JAMES L. BONTA, THE LEVEL OF SERVICE INVENTORY-REVISED USER'S MANUAL 1 (2001)).
22. I have adopted the term "off label" from the medical context, in which it is used to describe the practice of prescribing a medication to treat a condition other than the condition for which it has gained FDA approval. See U.S. FOOD & DRUG ADMIN., Understanding Unapproved Use of Approved Drugs "Off Label," https://www.fda.gov/ForPatients/Other/ OffLabel/ucm20041767.htm [https://perma.cc/ 5EDC-HN5T] (defining "off-label"). See generally Christopher M. Wittich et al., Ten Common Questions (and Their Answers) About Off-Label Drug Use, 87 MAYO CLINIC PROC. 982 (2012)
(describing the practice of prescribing medicine in "off-label" ways); see also Sandra Mayson, Off-Label Law Enforcement 1 (2018) (unpublished manuscript) (on file with author) (work-in-progress using the term 'off label' to describe the use of coercive force for a purpose "other than [the one] it was ostensibly designed to serve"). In their recent white paper, Hyatt and Chanenson offer a typology of ways jurisdictions use risk assessment tools at sentencing and characterize certain uses as "off label."
See Hyatt & Chanenson, supra note 19, at 7-8 (distinguishing between jurisdictions that systemically integrate risk assessment information from those that use the information in "off label" ways). In contrast to Hyatt and Chanenson, however, this Article contends that all uses of actuarial risk assessment methods at sentencing are "off label" uses.
This Article reexamines the practice and promises of actuarial sentencing in light of this observation. It scrutinizes the meaning of "risk" in the correctional and sentencing contexts, the different mechanisms available for responding to risk in these distinct contexts, and the consequences of those decisions. This anal-ysis reveals that integrating actuarial risk assessment tools into sentencing deci-sions departs from their original design and purpose. It identifies three points of departure: the decisions actuarial risk assessment tools are used to support, the punishment theories that support those decisions, and the consequences of those decisions.
Ultimately, the Article contends that the application of actuarial risk assessment information to sentencing decisions is neither a simple matter of "follow[ing] the evidence"2 3 to smarter sentencing decisions, nor is it inconsequential. Rather, this "off label" use can lead to negative, unintended consequences: it can justify an increase-rather than a decrease-in our use of incarceration, and in a way that undermines the fairness and integrity of our criminal justice system.
The Article proceeds as follows. Part I describes the practice and promises of actuarial sentencing. It identifies the recent emergence of two sentencing-specific applications of actuarial risk information and explores why this practice has gained such enthusiastic support. Part II identifies an as-yet unexamined reason to question the laudatory claims of actuarial sentencing proponents: these tools are not designed for use in the sentencing context. To bring this critique into focus, Part II offers a historical account of the origins of actuarial risk prediction that explores the differences between how and why these tools emerged, and how and why they are integrated into sentencing. Finally, Part II reveals that this repurposing of actuarial information leads to an unintended-or at least unac-knowledged-punishment consequence, which is antithetical to the tools' reha-bilitative origins.
Part III identifies the "off label," unintended consequences of incorporating actuarial information into sentencing decisions. Specifically, Part III contends that actuarial sentencing can lead to more-not less-reliance on incarceration. This counterintuitive possibility results from the way actuarial tools define and measure risk. They define risk broadly, as the likelihood that an individual will commit, or be arrested for, an offense of unspecified severity, and measure it based on a range of characteristics-such as gender and educational history-that are anathema to a just sentencing inquiry. Perhaps this prediction mechanism is acceptable, or at least understandable, when the tools are used for their intended Recently-in another arguably "off-label" application-actuarial risk information has been used to guide bail decisions. That use has been the subject of increasing scholarly scrutiny. See, e.g., Lauryn P. Gouldin, Defining Flight Risk, 85 U. CHI. L. REv. 677 (2018); Sandra G. Mayson, Dangerous Defendants, 127 YALE L.J. 490 (2018). This paper focuses exclusively on the issues that arise when risk assessment is used for punishment purposes. As the bail/pre-trial detention decision is not considered punishment, see U.S. v. Salerno, 481 U.S. 739, 748 (1987), it involves a risk inquiry that is distinct from sentencing. Therefore, it is beyond the scope of this paper.
23. See Hyatt et al., supra note 3, at 267 (urging the United States Sentencing Commission to "follow the evidence" by integrating risk assessment into sentencing).
purpose-to administer punishment in a way that reduces an individual's risk of recidivism. But when this mechanism is applied at sentencing, it imposes an underappreciated systemic risk that cannot be countenanced: that we will incar-cerate someone who poses no risk to public safety because of non-culpable, per-sonal characteristics.
I. THE RISE OF ACTUARIAL SENTENCING
Around 2000, a handful of jurisdictions began providing risk assessment infor-mation to judges in advance of sentencing.24 Over the last decade, and particu-larly over the last few years, actuarial sentencing has gained the support of a number of influential organizations and policymakers, including the American Law Institute,25 the National Center for State Courts (NCSC),26 the Conference of Chief Justices and the Conference of State Court Administrators,2 7 and current and former state sentencing commissioners.2 This Part seeks to explain this en-thusiastic embrace of actuarial sentencing before turning, in Parts II and III, to an analysis of why this enthusiasm is unwarranted.
This Part begins by providing an overview of the practice of actuarial sentenc-ing. This overview first explains what actuarial tools are and how they are used, and then describes how these tools are used in the sentencing context. It identifies two sentencing-specific applications of actuarial risk information: decisions about sentence length and determinations of sentence location. The Part then surveys the literature supporting this new practice to identify the purported promises of actuarial sentencing.
A. THE PRACTICE OF ACTUARIAL SENTENCING
Actuarial sentencing is the practice of using actuarial risk assessment tools to guide sentencing decisions. It consists of two components: the actuarial assess-ment of risk and the integration of that assessassess-ment into sentencing decisions. This section describes each of these components in turn.
1. Actuarial Risk Assessment
Actuarial risk assessment tools are surveys that guide the inquirer through a se-ries of questions about the presence or absence of recidivism risk factors in the subject.29 The inquirer allocates the subject a certain number of points for each factor, each of which is weighted according to the strength of its correlation with
24. See Warren, supra note 11, at 157 (noting Iowa has included risk assessment information in presentence reports since 2000 and Missouri began sending such information to judges in 2005). As will be discussed in further detail below, Virginia was an early adopter of actuarial risk assessment sentencing, initiating a pilot program in 1994 to use risk information to change an offender's sentence.
See infra notes 71-74 and accompanying text.
25. See MODEL PENAL CODE: SENTENCING § 6B.09 (AM. LAW INST., Proposed Final Draft 2017).
26. See CASEY ET AL., supra note 8, at 7.
27. See CONFERENCE OF CHIEF JUSTICES & CONFERENCE OF STATE COURT ADM'RS, supra note 9.
28. See supranote 10.
29. See, e.g., 42 PA. CONS. STAT. § 2154.7(e) (2010) (defining "risk assessment instrument" as "an empirically based worksheet which uses factors that are relevant in predicting recidivism").
recidivism.30 Based on the total score, the subject is identified as having a "low," "moderate," or "high" risk of recidivism.31
Thus, in the context of the criminal justice system, actuarial risk assessment is the process of using characteristics that statistically correlate with recidivism ("recidivism risk factors") to predict "'who will or will not behave criminally' in the future.'
"32
Actuarial risk assessment tools vary in the number and type of recidivism risk factors they consider.33 Despite the existence of a variety of tools, however, each tool considers criminal history34 and other factors. Their algorithm can thus be distilled down to "criminal history plus" other characteristics, and the plus varies from tool to tool. Some of these plus factors are "static," which means they are statistically "related to recidivism but cannot be altered through the delivery of services," such as age at first offense, gender, and family criminality.35 Others are "dynamic," or can change through the passage of time or intervention, such as current age, employment status, antisocial attitude, and substance addiction.36
Although all of the tools aim to measure the individual's risk of recidivism, there is some variance in how they define recidivism. Some tools, for example, define recidivism to include rearrest for an offense of any severity, regardless of whether the arrest leads to filing of criminal charges or a conviction. Those tools, therefore, predict the likelihood someone with the defendant's characteristics will be rearrested
30. See NATHAN JAMES, CONG. RESEARCH SERV., R44087, RISK AND NEEDS ASSESSMENT IN THE CRIMINAL JUSTICE SYSTEM 2 (2015) (describing an overview of the risk assessment process); Edward J. Latessa & Brian Lovins, The Role of Offender Risk Assessment: A Policy Maker Guide, 5 VICTIMS & OFFENDERS 203, 206, 210-12 (2010) (same). For some instruments, the weight of the particular factor is apparent from the face of the tool or survey. See id. at 210-11. Some instruments, such as COMPAS, use a proprietary algorithm, and the accounting mechanism is not publicly available. See State v. Loomis, 881 N.W.2d 749, 761 (Wis. 2016) (explaining that the company that developed the COMPAS instrument "considers COMPAS a proprietary instrument and a trade secret" and therefore the company "does not disclose how the risk scores are determined or how the factors are weighed").
31. JAMES, supra note 30, at 2.
32. MODEL PENAL CODE: SENTENCING § 6B.09 rptrs. note a (AM. LAW INST., Proposed Final Draft 2017) (quoting Stephen D. Gottfredson & Laura J. Moriarty, Statistical Risk Assessment: Old Problems and New Applications, 52 CRIME & DELINQ. 178, 192 (2006)).
33. See Eaglin, supra note 1, at 81 (describing different tools).
34. JULIAN V. ROBERTS, PUNISHING PERSISTENT OFFENDERS: EXPLORING COMMUNITY AND OFFENDER PERSPECTIVES 22 (2008) (identifying criminal history as a "central component of the most common" risk assessment tools, such as the LSI-R or the salient Factor Score); John Monahan & Jennifer L. Skeem, Risk Assessment in Criminal Sentencing, 12 ANN. REV. CLINICAL PSYCHOL. 489, 503 (2016) (noting that "[i]t has long been axiomatic in the field of risk assessment that past crime is the best predictor of future crime" and therefore, "[a]ll actuarial risk assessment instruments reflect this empirical truism. The empirically derived California Static Risk Assessment Instrument, for example, contains 22 risk factors for criminal recidivism, fully 20 of which-all but gender and age-are indices of past crime").
35. JENNIFER K. ELEK ET AL., NAT'L CTR. FOR STATE COURTS, USING RISK AND NEEDS ASSESSMENT INFORMATION AT SENTENCING: OBSERVATIONS FROM TEN JURISDICTIONS 1 (2015); see also Latessa & Lovins, supra note 30, at 208-09.
36. See Latessa & Lovins, supra note 30, at 209; see also ELEK ET AL., supra note 35, at 46.
37. See Nancy J. King, Sentencing and Prior Convictions: The Past, the Future, and the End of the Prior-Conviction Exception to Apprendi, 97 MARQ. L. REV. 523, 544 (2014) (noting that recidivism measures may "include any subsequent violent crime, felony arrest, felony conviction, conviction for any crime including a misdemeanor, or violation of a condition of supervised release"). For a critique of
within the specified time frame, without reference to actual culpability. Others define recidivism as reconviction for a new offense, but there is variation even among this subset of tools. Some define recidivism as reconviction for any offense, including a violation of probation; others define recidivism more narrowly as only reconviction for a felony.38
Regardless of how any particular tool defines recidivism, a few common prin-ciples unite all of them. First, the prediction remains actuarial.39 The risk assess-ment process is "about predicting group behavior (identifying groups of higher risk offenders). It is not about prediction at the individual level."4 That is, risk assessment tools identify groups of high-risk offenders, not a particular high-risk individual.41 The risk score indicates the likelihood someone who shares an indi-vidual's characteristics will recidivate, not the likelihood that particular a individ-ual will recidivate.
Second, the risk score indicates the likelihood of whether someone with the offender's characteristics will recidivate, not how they are likely to recidivate. Indeed, most tools do not distinguish between the likelihood the offender will recidivate by committing a serious offense or a low-level drug or property crime, but rather provide only a general prediction of the likelihood of recidivism.4 2 One exception is the Correctional Offender Management Profiling for Alternative Sanctions (COMPAS) tool, which distinguishes between "General Recidivism Risk" and "Violent Recidivism Risk. 43
Finally, the risk score indicates only the likelihood of future behavior; it does not indicate the relative severity of the crime the subject has most recently
defining recidivism to include arrest, see Anna Roberts, Arrests as Guilt (May 17, 2018) (unpublished manuscript) (on file with author).
38. See Kelly Hannah-Moffat, Actuarial Sentencing: An "Unsettled" Proposition, 30 JUST. Q. 270, 278-79 (2013) ("Among the available risk tools, recidivism is variably defined as re-arrest,
reconviction, or re-incarceration .... Many actuarial risk instruments do not differentiate between types of recidivism."); see also King, supra note 37, at 544 n.109 (citing additional sources).
The definition of recidivism may even vary within a particular jurisdiction. Virginia, for example, uses actuarial risk assessment information to guide decisions about whether to divert low-level offenders from prison and increase the sentencing guidelines recommendation for certain sex offenders. The diversionary risk assessment tool defines recidivism as a subsequent conviction for a felony offense within three years. See VA. CRIMINAL SENTENCING COMM'N, 2014 ANNUAL REPORT 94 (2014). The sex offender risk assessment tool, by contrast, defines recidivism as rearrest "for a new sex offense or other crime against a person." See id. at 39.
39. See CHRISTOPHER SLOBOGIN, PROVING THE UNPROVABLE: THE ROLE OF LAW, SCIENCE, AND SPECULATION IN ADJUDICATING CULPABILITY AND DANGEROUSNESS 101 (2007) ("An actuarial approach relies, as insurance actuaries do, on a finite number of preidentified variables that statistically correlate to risk and that produce a definitive probability or probability range of risk.").
40. Wis. DEP'T OF CORR., ELECTRONIC CASE REFERENCE MANUAL: COMPAS ASSESSMENT FREQUENTLY ASKED QUESTIONS, https://doc.helpdocsonline.com/dcc-business-process [https://perma.
cc/NYA3-2ZZ8]. 41. Id.
42. See Hannah-Moffat, supra note 38, at 278-79 (noting that "[m]any actuarial risk instruments do not differentiate between types of recidivism").
43. See NORTHPOINTE, PRACTITIONERS GUIDE TO COMPAS 1 (2012), http://www.northpointeinc.
com/files/technical documents/FieldGuide2_081412.pdf [https://perma.cc/K9H6-N29S] (describing the
committed. As Latessa and Lovins explain, "though a felon has been con-victed of a more serious offense than a misdemeanant, their relative risk of reoffending may have nothing to do with the seriousness of the crime."44 In fact, most risk assessment instruments do not incorporate the crime of con-viction into the prediction inquiry,45 as that factor has not been found to be predictive of recidivism.46
2. Sentencing Actuarially
Surprisingly, most jurisdictions that encourage or require actuarial sentencing do not restrict how judges may use this information at sentencing. Instead, they simply direct the department responsible for preparing the presentence report to conduct a risk assessment and provide it to the judge, without guidance as to the purposes for which it may be used.47
It appears that, until recently, judges who were authorized to consider risk information were doing so to identify and impose conditions of probation.4 8
An analysis of the Model Penal Code's new section on evidence-based sen-tencing,4 9 along with state case law and sentencing guidelines, reveals that two sentencing-specific actuarial practices have emerged: determinations about sen-tence length and determinations about sensen-tence location. Each will be explored below.
These applications are sentencing-specific because only a sentencing judge is empowered to make these decisions. The power to determine the severity of a sentence-to determine how much punishment is due a particular offender for a particular offense-is a core judicial function.0 After the judge imposes a sen-tence, the responsibility to execute that sentence then shifts to the executive branch, specifically correctional authorities and parole boards (if the jurisdiction allows for parole).5 1 Neither of these institutional actors, however, can reverse or
44. Latessa & Lovins, supra note 30, at 205-06.
45. See Starr, supra note 1, at 811 (noting that "almost none" of the actuarial risk instruments she
studied "include the crime for which the defendant was convicted in the case at hand").
46. See JAMES, supra note 30, at 7 ("The seriousness of the current offense is not a risk factor.").
47. See, e.g., WASH. REV. CODE § 9.94A.500(1) (2014); see also Starr, supra note 1, at 839-40 (noting that "[g]enerally . . .risk predictions are simply provided to sentencing judges and parole boards").
48. See Warren, supra note 11, at 157 (finding that the ten jurisdictions that had incorporated risk assessment into sentencing by 2010 used it "primarily for the purpose of determining the conditions of probation supervision").
49. MODEL PENAL CODE: SENTENCING § 6B.09 (AM. LAW INST., Proposed Final Draft 2017). 50. Commonwealth v. Cole, 10 N.E.3d 1081, 1089 (Mass. 2014) ("At the core of the judicial function is the power to impose a sentence."); Commonwealth v. Rodriguez, 962 N.E.2d 711, 718 (Mass. 2012) (describing the power to sentence as "a quintessential judicial power").
51. Cole, 10 N.E.3d at 1089 ("Once a sentence is imposed, the executive branch holds the power and responsibility of executing it."). Jurisdictions differ as to whether probation authorities are judicial or executive branch actors. See NAT'L CTR. FOR STATE COURTS, CCJ AND COSCA SURVEY OF EVIDENCE-BASED PRACTICES IN SENTENCING AND PROBATION: BRANCH RESPONSIBLE FOR PROBATION 1 (2012), http:// www.ncsc.org/-/media/microsites/files/csi/branch-responsible-for-probation.ashx [https://perma.cc/8UB
S-NE8V] (finding that in some jurisdictions probation is a function of the executive branch, in some it is a judicial branch function, and in others it is a mixed function).
modify the judge's sentencing decision. Even when a parole board decides to release a defendant from incarceration before the expiration of his sentence, for example, that decision does not and cannot change the sentencing judge's deci-sion about the length of his sentence. The defendant must complete her sentence, albeit under community rather than institutional supervision.2
a. Sentence Length Decisions
The length of a defendant's sentence-how long he or she will be under the supervision and control of state correctional authorities-is a paradigmatic sen-tencing decision. This decision is entrusted solely to the sensen-tencing judge, subject to the parameters set by the state or federal legislature and recommendations of any relevant sentencing commission. 3
The recently approved Model Penal Code endorses the use of risk assessment information to inform this key sentencing decision. Model Penal Code § 6B.09(2) orders a sentencing commission to develop actuarial instruments that "estimate the relative risks that individual offenders pose to public safety through their future criminal conduct" and "incorporate [this information] into the sentencing guide-lines."54 The commentary clarifies that this provision "would permit the use of actuarial offender risk assessments as a basis for punishments more severe than offenders would otherwise have received."5 5 Meanwhile, § 6B.09(3) directs that "unusually low-risk" offenders should be sentenced to community sanction or a "shorter prison term" than otherwise required by the statute or guidelines.5 6
Virginia, an early adopter of actuarial sentencing practices, was the first to employ actuarial risk assessment to sentence length decisions.5 7 Since 2001, it has authorized the use of actuarial information to increase sentencing exposure for adults who are convicted of certain sex offenses and predicted to pose a high risk of recidivism. 8 Recently, other jurisdictions have begun to follow suit. In July 2017, for example, Kansas began using risk assessment results to set the sen-tencing parameters for juvenile offenders. Juveniles convicted of a felony who score as a low or moderate risk may be sentenced to up to fifteen months, whereas
52. As the Supreme Judicial Court of Massachusetts recently explained, the parole board's authority is "limited to the release from custody of a defendant within the maximum term of imprisonment imposed by the sentencing judge." Cole, 10 N.E.3d at 1087. It "has the power only to permit a defendant to serve the balance of his term of imprisonment outside the prison walls,... and the power to revoke the parole permit and return the defendant to prison or jail for the balance of his term of imprisonment."
Id.
53. See supra note 50.
54. MODEL PENAL CODE: SENTENCING § 6B.09(2) (AM. LAW INST., Proposed Final Draft 2017).
55. Id. § 6B.09 cmt. e. 56. Id. § 6B.09(3).
57. See Jonathan Simon, Reversal of Fortune: The Resurgence of Individual Risk Assessment in Criminal Justice, 1 ANN. REV. L. & Soc. Sci. 397, 407 (2005) (noting in 2005 that Virginia was the only
state to use actuarial risk assessment to identify "high rate" offenders).
58. VA. CRIMINAL SENTENCING COMM'N, supra note 38, at 39. Virginia began using risk prediction to
authorize an increase in sentencing exposure for sex offenders deemed to be a high risk of recidivism in
those who score as a high risk may be sentenced up to eighteen months.5 9
As mentioned above, however, most jurisdictions that authorize actuarial sen-tencing simply direct that the risk information be provided to a judge in advance of sentencing without any guidance on how it should-or should not-be used.60 Washington State, for example, simply advises that a sentencing court "may order the department [of probation] to compete a risk assessment report," and that if that report is "available before sentencing, the report shall be provided to the court."6 1 Apparently, then, judges in these jurisdictions may use the results for any purpose they deem appropriate, including to inform their decision about the appropriate length of a sentence (subject to any applicable statutory and constitu-tional constraints). Cognitive behavioral research into the anchoring effect62 sug-gests judges who receive predictive risk information may modify their sentence
in the direction of the risk prediction.63 A few empirical studies support this
inference.64
This common-sense inference about the influential power of risk assessment information has been demonstrated in at least one actual case. In Wisconsin, risk assessment predictions are provided to the judge along with the presentence investigation report.65 In 2013, Paul Zilly pled guilty to stealing a lawnmower in Wisconsin, and the prosecutor recommended one year in jail followed by supervi-sion. 66 At sentencing, however, after noting Mr. Zilly scored as high-risk for vio-lent recidivism and medium-risk for general recidivism, which the judge
59. See KAN. SENTENCING COMM'N, KANSAS SENTENCING GUIDELINES DESK REFERENCE MANUAL 7 (2016) (describing changes to the law regarding case length limits, effective July 1, 2017). The court may also use the risk information to set the juvenile's term of probation, and may extend the term of a high-risk juvenile's term of probation by up to six months "if a juvenile needs time to complete an evidence-based program determined to be necessary based on the results of a validated risk and needs assessment." Id. at 9-10.
60. See supra notes 47-48 and accompanying text. 61. WASH. REV. CODE § 9.94A.500(1) (2014).
62. See Fritz Strack & Thomas Mussweiler, Explaining the Enigmatic Anchoring Effect: Mechanisms of Selective Accessibility, 73 J. PERSONALITY & SOC. PSYCHOL. 437, 437 (1997) (defining
the "anchoring effect" as "a biased estimate toward an arbitrary value considered by judges before making a numerical estimate").
63. See Starr, supra note 1, at 867 ("Even if a particular judge does not really trust the instrument, its prediction might influence her thinking through anchoring.").
64. For example, Professor Sonja Starr conducted an informal study with eighty-three criminal law students and found, similarly, that prediction information influenced participants' decisions about an appropriate sentence for a hypothetical case. See id. at 867-69 (describing the study). A study conducted by the Pennsylvania Sentencing Commission also suggests that judges, probation officers, and criminal law practitioners are influenced by exposure to risk assessment information. See PA. COMM'N ON SENTENCING, RISK/NEEDS ASSESSMENT PROJECT, INTERIM REPORT 8: COMMUNICATING RISK AT
SENTENCING 1, 6-11 (2014), http://www.hominid.psu.edu/specialty-programs/pacs/publications-and-research/risk-assessment/phase-i-reports/interim-report- 8-communicating-risk-at-sentencing/view [https:// perma.cc/NY63-TK3Q] (describing the study and its findings).
65. See State v. Loomis, 881 N.W.2d 749, 754 (Wis. 2016).
66. See Julia Angwin et al., Machine Bias: There's Software Used Across the Country to Predict Future Criminals. And It's Biased Against Blacks., PROPUBLICA (May 23, 2016), https://www.propublica.org/ article/machine-bias-risk-assessments-in-criminal-sentencing [https://perma.cc/RA73-X8SW] (discussing the Zilly case).
characterized as "about as bad as it could be," the judge rejected the plea deal and instead imposed a sentence of two years in prison followed by three years of supervision.67 After Mr. Zilly appealed, the court reduced his sentence to eighteen months, admitting, "Had I not had the [risk assessment information], I believe it would likely be that I would have given one year, six months."6 8
b. Sentence Location Decisions
A second paradigmatic sentencing decision is the determination of where the defendant will serve his sentence: in institutional confinement (prison or jail), under community supervision (probation), or a combination of the two. This is a sentencing-specific decision because only the sentencing judge is empowered to determine the initial, default location of a defendant's sentence.6 9 Certainly, cor-rections officials can alter this location, based on developments that occur after the sentence is determined. A defendant sentenced to community supervision who does not comply with the terms of probation may be incarcerated, in some instances without judicial intervention,70 and a defendant sentenced to incarcera-tion may be released into the community to serve the remainder of her sentence on parole, again without the approval of a judge.71 Significantly, however, all of these decisions are made after sentence has been announced, and, absent subse-quent intervening developments, the defendant will serve her sentence in the location set by the judge.
In contrast to the sentence-length use discussed above, which remains some-what uncommon, the use of actuarial risk information to inform sentencing-loca-tion decisions is fairly prevalent. Jurisdicsentencing-loca-tions have integrated risk predicsentencing-loca-tions into at least three different sentence-location decisions: (1) whether to sentence a defendant to probation or incarceration, (2) whether to divert otherwise prison-bound offenders to jail or probation, and (3) whether to suspend part or all of a prison sentence for one spent in the community.
First, some jurisdictions use actuarial information to determine whether a de-fendant who is eligible to serve his sentence in the community or at an institution should be sentenced to probation or incarceration.7 2 The 2016 Wisconsin case of
State v. Loomis illustrates this approach.73 Eric Loomis was arrested in Wisconsin while driving a stolen car. The state alleged he was the driver in a drive-by shoot-ing; however, Mr. Loomis denied involvement in the shooting and maintained he
67. Id. 68. Id.
69. Again, as with the length of the defendant's sentence, this decision is restrained by any relevant sentencing statutes or guidelines.
70. See Fiona Doherty, Obey All Laws and Be Good: Probation and the Meaning of Recidivism, 104 GEo. L.J. 291, 326 (2016) (noting that some states "have allowed probation officers to impose short jail or prison sentences as administrative sanctions without specific court approval" and providing examples).
71. See Commonwealth v. Cole, 10 N.E.3d 1081, 1089 (Mass. 2014) (noting that "the judiciary may not interfere" with parole decisions).
72. See, e.g., State v. Loomis, 881 N.W.2d 749 (Wis. 2016). 73. Id.
drove the car only after the shooting occurred. He ultimately pleaded guilty to operating a motor vehicle without consent and to eluding a police officer.7 4
The presentence investigation report submitted to the judge included a Correctional Offender Profiling for Alternative Sanctions (COMPAS) risk assess-ment, which indicated Mr. Loomis presented a high risk of pretrial, general, and violent recidivism.7 In ruling out probation-that is, in deciding to incarcerate Mr. Loomis instead of sentencing him to a community-based sentence-the cir-cuit court noted that Mr. Loomis had been "identified, through the COMPAS assessment, as an individual who is at high risk to the community," and that one reason it was ruling out probation was that "the risk assessment tools that have been utilized, suggest that you're extremely high risk to re-offend. 76
In jurisdictions that follow this first approach, actuarial information can weigh in favor of either a community-based (probation) or institutional (prison or jail) sentence-risk prediction may lead to either a better or worse outcome for the de-fendant. Under a second approach, however, actuarial risk predictions can lead only to a better location decision for the defendant. This diversionary approach uses actuarial information to identify low-risk, prison-bound defendants and sen-tence them to community supervision or jail (meaning a sensen-tence less than twelve months) in lieu of prison.77 Defendants in these jurisdictions would, absent consideration of risk assessment information, be sentenced to prison.78 The Model Penal Code, for example, instructs the sentencing commission to "develop actuarial instruments or processes to identify offenders who ... are subject to a presumptive or mandatory sentence of imprisonment" but present an "unusually low risk to public safety," and recommends that the sentencing judge have discretion to sentence such offenders to a "community sanction rather than a prison term. '71
Virginia employs risk information in precisely this way.8 It uses a "nonviolent offender risk assessment instrument" to actuarially predict the recidivism risk of defendants who have been convicted of specified nonviolent felonies and who, according to state sentencing guidelines, are recommended for a prison
74. Id. at 754. 75. Id. at 754-55.
76. Id. at 755 (quoting the circuit court at sentencing).
77. See, e.g., VA. CRIMINAL SENTENCING COMM'N, supra note 38, at 36 ("The goal of the nonviolent
risk assessment instrument is to divert low-risk offenders who are recommended for incarceration on the guidelines to an alternative sanction other than prison or jail. Therefore, nonviolent offenders who are recommended for probation/no incarceration on the guidelines are not eligible for the assessment.").
78. See id.
79. MODEL PENAL CODE: SENTENCING § 6B.09(3) (AM. LAW INST., Proposed Final Draft 2017). And, as discussed above, the Model Penal Code also authorizes imposing shorter sentences on these "unusually low-risk" offenders. Id.; see supra note 56 and accompanying text.
80. Virginia was a very early adopter of actuarial sentencing practices. See MODEL PENAL CODE:
SENTENCING § 6B.09 rptrs. note d (AM. LAW INST., Proposed Final Draft 2017) (describing Virginia as "the first state to develop an actuarial risk-assessment tool to be used at sentencing for purposes of diverting low-risk offenders otherwise bound for prison into community sanctions"). Virginia has incorporated risk assessment into incarceration decisions statewide since 2002. VA. CRIMINAL
sentence.81 This tool differs from the sex offender risk assessment tool discussed above.8 2 Virginia's current nonviolent drug offender risk assessment tool consid-ers a handful of factors deemed statistically significant to recidivism prediction, including: the defendant's gender and age at the time of the offense; prior juve-nile adjudications or adult felony convictions; and whether the defendant was arrested or confined within the twelve months leading up to the offense.83 And until 2013, the tool also considered marital status and employment status.84
Offenders who score below a certain threshold are recommended for an alter-native sentence. Under this regime, an alteralter-native sentence includes "anything short of an actual state prison sentence," such as probation supervision, commu-nity service, day or evening reporting, drug treatment programs, and incarceration at a local correctional facility.86 Notably, a defendant is considered "diverted" if the guidelines recommended prison and she is sentenced instead to jail.8 Indeed, in Virginia, 2011, one of the most common "alternative punishments" meted under this program was a jail sentence instead of a prison sentence.88 Two im-portant observations flow from this insight. First, the "alternative" sentence was a diversion from prison, but not from incarceration. Those who were diverted to a jail sentence were nevertheless incarcerated, albeit at a local facil-ity instead of a state facilfacil-ity. Second, this practice demonstrates that there can be overlap between sentence-location and sentence-length decisions. By decid-ing to divert a defendant to jail instead of prison, Virginia judges routinely decided to impose a shorter sentence than they would have without the risk assessment information.
A third method of integrating actuarial information into sentence-location decisions is to use it to decide whether to suspend part of a defendant's
81. VA. CRIMINAL SENTENCING COMM'N, supra note 38, at 86; see also Brian J. Ostrom & Neal B. Kauder, The Evolution of Offender Risk Assessment in Virginia, 25 FED. SENT'G REP. 161, 166 (2013).
Certain individuals within these offense categories are categorically excluded from consideration for risk-based diversion. For example, individuals with a prior felony conviction for a violent offense or those convicted of selling one ounce or more of cocaine may not be considered for risk assessment diversion. VA. CRIMINAL SENTENCING COMM'N, supra note 38, at 86.
82. See supra note 58 and accompanying text.
83. See VA. CRIMINAL SENTENCING COMM'N, SENTENCING GUIDELINES COVERSHEET & WORKSHEET:
DRUG/OTHER, SECTION D (2017), http://www.vcsc.virginia.gov/worksheets 2017/DRG othws.pdf; see also BRIAN J. OSTROM ET AL., OFFENDER RISK ASSESSMENT IN VIRGIIA: A THREE-STAGE EVALUATION 12 (2002), http://www.vcsc.virginia.gov/risk-off rpt.pdf. This tool was designed precisely for this purpose and engages only in risk identification. In designing the tool, the Virginia Criminal Sentencing Commission decided that "needs assessment"-or what I have above characterized as risk intervention-was the work of probation officers and the Department of Corrections, not judges, and therefore intervention-was not encompassed in the risk tool. Ostrom & Kauder, supra note 81, at 166.
84. See OSTROM ET AL., supra note 83, at 12, 27; see also Ostrom & Kauder, supra note 81, at 167 n.4.
85. VA. CRIMINAL SENTENCING COMM'N, supra note 38, at 87. In fiscal year 2014, 47.5% of those assessed were recommended for an alternative sanction and almost 38% of those who were recommended received an alternative sentence. Id.
86. Ostrom & Kauder, supra note 81, at 162-63. 87. Id.
incarceratory sentence and replace that suspended portion with community super-vision. California recently adopted this approach. Since 2015, California Rules of Court have allowed courts to consider risk assessment information in determining the length and conditions of an individual's period of mandatory supervision.89 Mandatory supervision, like probation, is a period of supervised release in the community.9" However, "[m]andatory supervision . . . is not probation."9 1 Whereas probation is a period of community supervision that replaces a period of incarceration, mandatory supervision is a period of community supervision that follows incarceration. When sentencing individuals convicted of specified low-level crimes, California courts "must suspend execution of a concluding portion" of the sentence "as a period of mandatory supervision. "92 Notably, however,
California judges do not consider imposing a split sentence until they have al-ready ruled out a sentence of probation.9 3
B. THE PROMISES OF ACTUARIAL SENTENCING
Proponents of actuarial sentencing make three primary claims about the bene-fits of actuarial sentencing. First, they package actuarial sentencing as part of a systemic reorientation of the criminal justice system that will increase public safety by reducing recidivism.94 Second, they claim it simply increases the accu-racy of decisions judges are already making. Finally, they emphasize the benefi-ciaries of this innovation: the public, who save money while avoiding future victimization and the defendants who avoid incarceration.96 This section identi-fies and explores each of these promises.
1. Reducing Recidivism Through Rehabilitation
Many criminal justice reform advocates, including many actuarial sentencing proponents, identify the systemic embrace of retributivism, and the concomitant rise of a tough-on-crime ethos and harsh sentencing practices as the source of our current systemic dysfunction.97 As a result, they claim, we have a criminal justice
89. CAL. R. CT. 4.415(c).
90. See Chief Prob. Officers of Cal., Mandatory Supervision: The Benefits of Evidence Based Supervision Under Public Safety Realignmnet, 1 CPOC ISSUE BRiEF 1, 1 (2012) (defining mandatory supervision as "a court ordered period of time in the community under the supervision of the county probation department").
91. RICHARD CoUzENS ET AL., CRIMINAL JUSTICE REALIGNMENT FREQUENTLY ASKED QUESTIONS 5 (2014),
http://www.courts.ca.gov/partners/documents/cjr-faq.pdf ("Mandatory supervision [under California's criminal justice realignment] is not probation. Mandatory supervision may not be used until the judge denies probation and imposes a split sentence. The supervision is part of the sentence imposed by the
court.").
92. CAL. R. CT. 4.415(a).
93. See COUZENS ET AL., supra note 91.
94. See infra Section I.B.1.
95. See infra Section 1.B.2. 96. See infra Section 1.B.3.
97. See, e.g., Hon. William Ray Price Jr., Chief Justice of the Supreme Court of Mo., State of the Judiciary Address (Feb. 3, 2010), http://www.courts.mo.gov/page.jsp?id=36875 [https://perma.cc/ GM5M-WG3G] (critiquing "tough on crime" policies and arguing in favor of actuarial sentencing). See
system that resorts quickly to incarceration and, consequently, is struggling under the weight of mass incarceration. As retributivism is understood to be a contribut-ing cause of mass incarceration, reformers contend that we need a new "brand" of criminal justice that focuses on public safety instead of punishing past behavior.98
Actuarial sentencing is marketed as an important component of this new brand of public-safety focused criminal justice. For example, in 2011, Kentucky enacted a Public Safety and Offender Accountability Act that declared that the state's "primary objective of sentencing" is to "maintain public safety and hold offenders accountable while reducing recidivism and criminal behavior and improving outcomes for those offenders who are sentenced."99 In that same Act, Kentucky legislators declared that sentencing judges "shall consider . . .the results of a defendant's risk and needs assessment included in the presentence investigation." ' The Conference of Chief Justices and the Conference of State Court Administrators similarly linked actuarial sentencing to a new era of public-safety-focused sentencing law and policy. In their 2007 Resolution in support of actuarial sentencing practices, they "elevated recidivism reduction as an impor-tant consideration in the sentencing process" and stressed the importance of using risk assessment tools to reduce recidivism.0 1 And the NCSC's National Working Group on Using Risk and Needs Assessment Information at Sentencing identified the first benefit of actuarial sentencing as "[c]ontributing to public safety/avoid-ing further victimization by reducsafety/avoid-ing recidivism. 1 2
Actuarial sentencing proponents specify that this practice will actualize its recidivism reduction goal through rehabilitation. For example, the Indiana Court Times, a publication of the Indiana State Court system, claims that the use of risk assessment tools at sentencing will "enhance efforts to rehabilitate offenders, reduce recidivism, and increase public safety."1 3 And in his 2010 State of the Judiciary Address, Missouri Chief Justice William Ray Price Jr. called for a movement away from "anger-based sentencing that ignores cost and effective-ness" toward actuarial sentencing, which "assesses each offender's risk and then fits that offender with the cheapest and most effective rehabilitation that he or she needs."1 4 The NCSC's model curriculum for teaching judges about actuarial
generally Erin R. Collins, Status Courts, 105 GEO. L.J. 1481, 1513-17 (2017) (discussing the connection
between retributivist policies and mass incarceration). 98. See KELLY ET AL., supra note 7, at 177.
99. Ky. REV. STAT. ANN. § 532.007(1) (West 2011).
100. Id. § 532.007(3)(a); see also COUNCIL OF STATE Gov'TS, LESSONS FROM THE STATES: REDUCING RECIDIVISM AND CURBING CORRECTIONS COSTS THROUGH JUSTICE REINVESTMENT 4 (2013), https:// csgjusticecenter.org/wp-content/uploads/2013/04/FINAL State Lessons mbedit.pdf (noting that "many states fail to focus their incarceration . . .on the people most likely to commit future crimes" and advocating for the use of risk assessment instruments).
101. CASEY ET AL., supra note 8, at 3.
102. Id. at 7.
103. Indiana's New Risk Assessment Tools: What You Should Know, IND. CT. TIMES (Apr. 13, 2011), http://indianacourts.us/times/2011/04/risk-assessment/ [https://perma.cc/U375-W8ZW].
104. Price, supra note 97. Chief Judge Price also emphasized that actuarial sentencing would lead to diversion of some offenders from prison, and "remov[al of] others from prison more quickly-after they have learned their lesson, but before they are mined by worse offenders." Id.
sentencing defines the practice as one that is "based on 'corrections' principles... used to reduce recidivism."' The NCSC's National Working Group similarly links actuarial sentencing to rehabilitation. The Group dedicates five pages of its guide to why and how jurisdictions should implement actuarial sentencing to explicating the Risk-Needs-Responsivity (RNR) principle.10 6 As will be discussed in greater detail in Part II, a core tenet of the RNR principle is that recidivism risk should be identified so that it can be reduced through appropriate and effective re-habilitative programming.
Thus, one marketing promise of actuarial sentencing is that it is one of many reforms that will help increase public safety by reducing recidivism. Its propo-nents contend that actuarial tools improve sentencing by shifting the emphasis from retribution to rehabilitation-from extracting punishment for past bad behavior to increasing public safety by reforming offenders.
2. Refining Risk Prediction
A related, but somewhat conflicting, claim offered in favor of actuarial sen-tencing is that it simply enhances, but does not fundamentally change, the risk in-quiry judges are and should be making at sentencing. From this perspective, actuarial sentencing appears to be a simple matter of "follow[ing] the evidence" to a better sentencing decision.1 7 Its defense generally consists of three cumula-tive assertions: (1) that risk prediction is a normacumula-tively sound sentencing func-tion; (2) that these tools enhance the accuracy of this established predicfunc-tion; and (3) that its use therefore enhances sentencing decisions.
First, and often with a reference to Jurek v. Texas,1"8 actuarial sentencing propo-nents emphasize that prediction of future behavior is an established component of sentencing decisions.10 9 As the Supreme Court noted in Jurek, "any sentencing authority must predict a convicted person's probable future conduct when it engages in the process of determining what punishment to impose."' 0 Accordingly, the Model Penal Code reporters comment that adopting actuarial sentencing practices "recognizes that American sentencing systems will and should take account of an offender's future behavior, including the offender's amenability to rehabilitation and propensity to recidivate, when assigning penalties."' 1 Richard Kern, former
105. See NAT'L CTR. FOR STATE COURTS, EVIDENCE-BASED SENTENCING TO IMPROVE PUBLIC SAFETY & REDUCE RECIDIVISM: A MODEL CURRICULUM FOR JUDGES 10 (2009), http://www.ncsc.org/
-/media/Microsites/Files/CSI/Education/FacultyHandbook.ashx [https://perma.cc/GPA7 -3GZ6]. 106. See CASEY ET AL., supra note 8, at 4-8.
107. Hyatt et al., supra note 3, at 267.
108. 428 U.S. 262 (1976).
109. See, e.g., Christopher Slobogin, Risk Assessment, in THE OXFORD HANDBOOK OF SENTENCING AND CORRECTIONS 196, 203-05 (Joan Petersilia & Kevin R. Reitz eds., 2012) ("The Supreme Court, however, does not believe that risk assessment is antithetical to criminal justice. It has even approved death sentences based on dangerousness determinations (Jurek v. Texas 1976, 275-276).").
110. 428 U.S. at 274-76.
111. MODEL PENAL CODE: SENTENCING § 6B.09 cmt. a (AM. LAW INST., Proposed Final Draft 2017);
see also id. § 6B.09 cmt. e ("Judgments-or guesses-about offenders' future criminality have long
